科研之友
微信
新浪微博
Facebook
分享链接
摘要
This paper presents a unified social graph based text mining framework to identify digital evidences from chat logs data. It considers both users%26apos; conversation and interaction data in group-chats to discover overlapping users%26apos; interests and their social ties. The proposed framework applies n-gram technique in association with a self-customized hyperlink-induced topic search (HITS) algorithm to identify key-terms representing users%26apos; interests, key-users, and key-sessions. We propose a social graph generation technique to model users%26apos; interactions, where ties (edges) between a pair of users (nodes) are established only if they participate in at least one common group-chat session, and weights are assigned to the ties based on the degree of overlap in users%26apos; interests and interactions. Finally, we present three possible cyber-crime investigation scenarios and a user-group identification method for each of them. We present our experimental results on a data set comprising 1100 chat logs of 11,143 chat sessions continued over a period of 29 months from January 2010 to May 2012. Experimental results suggest that the proposed framework is able to identify key-terms, key-users, key-sessions, and user-groups from chat logs data, all of which are crucial for cyber-crime investigation. Though the chat logs are recovered from a single computer, it is very likely that the logs are collected from multiple computers in real scenario. In this case, logs collected from multiple computers can be combined together to generate more enriched social graph. However, our experiments show that the objectives can be achieved even with logs recovered from a single computer by using group-chats data to draw relationships between every pair of users.
- 出版日期2014-12
